Design of a high-accuracy air temperature measurement system using computational fluid dynamics and neural networks

Abstract

Global temperatures are rising by approximately 0.1 °C per decade. Existing air temperature measurement systems often report temperatures higher than actual air temperature due to the effects of solar radiation, leading to errors of up to 1 °C. As a result, there is an urgent need for a new temperature measurement system with improved radiation protection and ventilation capabilities. Furthermore, a specialized temperature error correction model is essential for the new system. Computational fluid dynamics (CFD) software was employed to simulate the radiation shielding and ventilation performance of the new system. Temperature differences between the new system and actual air temperature under various environmental conditions were quantified using CFD software. Subsequently, a specialized temperature difference correction model, incorporating multiple environmental variables, was developed using a neural network algorithm. Finally, the measurement accuracy of the new system was evaluated through field comparison experiments. During the experiments, a 076B fan aspirated temperature measurement system with an error of less than 0.03 °C served as the reference system. Before correction, the new system exhibited a maximum temperature difference of 0.69 °C and an average temperature difference of 0.35 °C compared to the reference system. The mean absolute error, root mean square error, and correlation coefficient between the temperature differences from the correction model and the experimental data were 0.07 °C, 0.08 °C, and 0.9 °C, respectively. After correction, the average temperature difference decreased to 0.06 °C. These results indicate that the new system has significant potential for high-accuracy temperature measurement.